Feature request

Feature description

To decouple CLI commands from node discovery latencies, we currently spawn a daemon process that caches the ROS graph, speeding up queries. But we do so sparsely. Several verbs implementations -- ros2 security generate policy, ros2 node info and ros2 param list, just to name a few -- first use the daemon via NodeStrategy and then use DirectNode, effectively racing against discovery (e.g. the node in the daemon may have discovered a remote node already, but the node created by the CLI may have not yet).

Thus, we should use the daemon correctly, potentially adding support for a larger subset of the ROS graph API in the process, and/or spin long enough so that discovery can take place.